Catalytic Reduction of CO2 to CO via Reverse Water Gas Shift Reaction: Recent Advances in the Design of Active and Selective Supported Metal Catalysts

The catalytic conversion of CO2 to CO via a reverse water gas shift (RWGS) reaction followed by well-established synthesis gas conversion technologies may provide a potential approach to convert CO2 to valuable chemicals and fuels. However, this reaction is mildly endothermic and competed by a strongly exothermic CO2 methanation reaction at low temperatures. Therefore, the improvement in the low-temperature activities and selectivity of the RWGS reaction is a key challenge for catalyst designs. We reviewed recent advances in the design strategies of supported metal catalysts for enhancing the activity of CO2 conversion and its selectivity to CO. These strategies include varying support, tuning metal–support interactions, adding reducible transition metal oxide promoters, forming bimetallic alloys, adding alkali metals, and enveloping metal particles. These advances suggest that enhancing CO2 adsorption and facilitating CO desorption are key factors to enhance CO2 conversion and CO selectivity. This short review may provide insights into future RWGS catalyst designs and optimization.


Introduction
The concentration of carbon dioxide (CO 2 ) in the atmosphere has increased dramatically, triggering a series of severe environmental problems, such as global warming, glacier melting, and ocean acidification [1][2][3][4][5]. These problems seriously threaten the human living environment. Therefore, CO 2 concentration in the atmosphere must be reduced. Many researchers have tried to solve these problems through carbon dioxide capture and storage (CCS) and carbon dioxide capture and utilization (CCU) [6][7][8][9][10]. However, CCS poses a risk of CO 2 leakage and causes unknown effects on the environment [11]. Thus, developing the CCU technology appears to be more meaningful than simply storing CO 2 [12,13].
The conversion of CO 2 to CO via a reverse water gas shift (RWGS) reaction [Eq. (1)] has been widely explored because the produced CO can be further converted to valuable chemicals and fuels through well-developed synthesis gas (CO and H 2 ) conversion technologies, such as Fischer-Tropsch (FT) synthesis and methanol synthesis. However, the RWGS reaction is competed by Sabatier reaction or CO 2 methanation [Eq. (2)] and methanol formation, resulting in a decreased CO yield [11,14,15]. Under atmospheric pressure, the yield of methanol is rather low or absent [16] and thus can be ignored. However, the strongly exothermic methanation reaction is thermodynamically more favored over the mildly endothermic RWGS reaction at low reaction temperatures [17][18][19][20][21]; as such, minimizing the methanation during RWGS becomes a great challenge. Therefore, active RWGS catalysts operated at low temperatures with enhanced CO selectivity and minimized CH 4 selectivity should be developed. CO may evolve as an intermediate in the hydrogenation of CO 2 to CH 4 , methanol, and other hydrocarbons [11,16,17,22,23], so understanding the mechanism of CO formation is also important for the rational design of catalysts for these processes.
Based on reaction kinetics, spectroscopy, and density functional theory (DFT) calculations, two mechanisms of RWGS have been proposed [24][25][26][27][28][29]. One is redox mechanism: CO 2 is reduced to CO with the catalyst being oxidized directly, and then H 2 reduces the catalyst again to complete the catalytic turnover [24]. This mechanism was originally proposed for Cu-based catalysts, where Cu is oxidized by CO 2 and reduced again by H 2 [24,25,30,31]. This mechanism has also been proposed for catalysts containing reducible oxides, which can be readily reduced and oxidized under reaction conditions [32][33][34][35][36].
Another is association mechanism: CO 2 adsorbs on the surface of catalysts and reacts with dissociated H to form an intermediate species, such as formate (*HCOO), carboxyl (*COOH), carbonate (CO 3 2− ), and bicarbonate (HCO 3 − ), which is decomposed to CO and H 2 O [27,37]. This mechanism was originally proposed for metal/reducible support catalysts [27,[38][39][40], such as Pt/CeO 2 . For this mechanism, the key intermediate proposed for the RWGS reaction may vary among different catalysts or the same catalyst operating under different conditions [41][42][43][44]. However, redox and association mechanisms have been debated. The mechanism of RWGS reaction is strongly dependent on catalyst types and reaction conditions; as such, no consensus has been reached.
Regardless of the mechanism of the RWGS reaction, an active catalyst requires two functionalities: (a) adsorption of CO 2 and disruption of one C-O bond of CO 2 ; (b) dissociation of H 2 and hydrogenation of O to form H 2 O. A selective RWGS catalyst needs a mild C-O dissociation ability and a weak CO adsorption ability to avoid the dissociation of CO and the further hydrogenation of CO to CH 4 and methanol. Therefore, the balance between C-O dissociation and hydrogenation is a key factor for the development of highly active and selective RWGS catalysts operating under mild conditions. The hydrogenation ability can be provided by a typical transition metal (e.g., Pt and Pd), whereas the C-O breaking ability can be achieved by a mildly oxophilic transition metal (e.g., Ru and Rh) [45][46][47] and may be further enhanced by a support or a promoter to improve CO 2 adsorption. The balance between hydrogenation and C-O bond dissociation abilities is evidently dependent on the properties and interactions of metals and supports (or promoters).
The development of the RWGS reaction has been briefly reviewed in several reviews focusing on the catalytic conversion of CO 2 to fuels and chemicals [12,[48][49][50]. For Porosoff et al. [48] reviewed various catalysts for the conversion of CO 2 to CO, methanol, and hydrocarbon fuels. Kattel et al. [51] reviewed the conversion of CO 2 to C 1 products of CO, methanol, and methane and focused on the role of the interface of metal/oxide catalysts in these reactions. Two reviews on the RWGS reaction have been conducted [12,37]. Daza and Kuhn [12] analyzed the feasibility of converting CO 2 to liquid fuels through the RWGS reaction. They also explored the supported metal catalysts and oxide catalysts for RWGS, and their reaction mechanisms. Su et al. [37] reviewed the reaction mechanism and more types of catalysts (supported metal catalysts, mixed oxide catalysts, and transition metal carbide catalysts) for the RWGS reaction.
In contrast to previous reviews providing a detailed overview of various types of catalysts for the RWGS reaction, this short review aims to present the recent advances in strategies developed to improve the selectivity and activity of a RWGS reaction. Although other types of catalysts have been widely studied [52][53][54], our review focuses on supported metal catalysts, which are the most intensively investigated catalyst for the RWGS reaction. In particular, this review considers the role of a support (or a promoter) and its interaction with transition metals in tuning the activity and selectivity of the RWGS reaction.

Strategies to Improve the Activity and Selectivity of the RWGS Reaction
Supported transition metal catalysts have been widely explored for CO 2 reduction. Depending on metal properties, Pd-and Pt-based catalysts are selective for producing CO, whereas catalysts based on oxophilic metals, such as Ni, Rh, and Ru, are more selective for producing CH 4 [55][56][57][58][59]. CH 4 can be a minor or major product under a RWGS reaction condition in the presence of these catalysts [55][56][57][58][59]. Thus, in addition to improving activities, tuning the selectivity of catalysts toward the RWGS reaction is important.
Transition metals, such as Pd and Pt, are highly active for activating H 2 . However, they weakly adsorb and activate CO 2 . Although DFT work has shown that CO 2 electrochemical reduction by H 2 on Pt (111) is possible [60], Kattel et al. [58] theoretically demonstrated that even a Pt nanoparticle (Pt 46 cluster) with low coordination surface sites is unable to catalyze the RWGS reaction because of a weak CO 2 adsorption. Kinetic Monte Carlo simulations have revealed that CO production is possible only when the CO 2 adsorption strength is further improved (Fig. 1a), and the weak adsorption of CO facilitates CO desorption while strong adsorption of CO produces more CH 4 (Fig. 1b) [58]. In addition, Kwak et al. [61] experimentally showed that even Pd atomically dispersed on a multiwall carbon nanotube (Pd/MWCNT) is 1 3 inactive (Table 1) for CO 2 reduction under mild conditions (< 500 °C) because neither Pd nor MWCNT can strongly adsorb and activate CO 2 under their reaction conditions. These works have indicated that bare transition metals are inefficient for the RWGS reaction because of their low abilities in CO 2 activation. Therefore, a support or a promoter and its interaction with a transition metal are essential for tuning the activity and selectivity of the RWGS reaction.

Varying Support
The reducibility and acid-base properties of a support can greatly influence reaction intermediates and CO 2 adsorption and activation. Hence, supports play an important role in the activity and selectivity of the RWGS reaction. Choosing an appropriate support is important to tune the activity and selectivity of this reaction. SiO 2 and Al 2 O 3 are irreducible supports. Their major role is to disperse the supported species, and they are not expected to directly activate CO 2 . Even though SiO 2 is considered an inert support, it can play a role in the RWGS reaction. As shown in Fig. 2a, CO 2 adsorbs at the interface of Pt/ SiO 2 ; in particular, C adsorbs on Pt, and O bonds with the hydroxyl of SiO 2 via hydrogen bonding [58]. The adsorption strength of CO 2 in the presence of SiO 2 enhances by 0.21 eV compared with that of a bare Pt nanoparticle, thus enabling the RWGS reaction [58]. Al 2 O 3 is another widely used irreducible support with weak acidity. In comparison with Pd/MWCNT, which cannot catalyze the RWGS reaction, 0.5% Pd atomically dispersed on Al 2 O 3 converts CO 2 to CO and yields CH 4 as a minor product (Table 1) [61]. Using operando diffuse reflectance infrared Fourier transformed spectroscopy (DRIFTS), Bobadilla et al. [62] suggested that CO 2 reacts with the hydroxyl of Al 2 O 3 to form bicarbonate species (Fig. 3a), which then reacts with H spilled over from Au to form formate. Formate eventually decomposes to CO. These studies have indicated that SiO 2 and Al 2 O 3 indirectly participate in the RWGS reaction by enhancing CO 2 adsorption. Normally, the activities of SiO 2 -and Al 2 O 3 -supported catalysts in a RWGS reaction are relatively low (Table 1), and they are mainly used to benchmark the activity of supported metals.
In contrast to SiO 2 and Al 2 O 3 , TiO 2 and CeO 2 are reducible supports, which readily generate oxygen vacancies, particularly in the presence of a metal to activate H 2 . An oxygen vacancy participates in a RWGS reaction through the strong adsorption of CO 2 and may directly activate CO 2 . Although different reaction pathways are proposed, the oxygen vacancy in proximity to a metal site plays a crucial role in enhancing the activity of the RWGS reaction.
The activity of metal catalysts supported on TiO 2 is normally higher in a RWGS reaction than that on an irreducible support. Kim et al. [32] conducted a kinetic study and proposed that a RWGS reaction follows a redox mechanism on Pt/TiO 2 catalysts. They performed H 2 temperature-programmed reduction and temperature-programmed desorption experiments and suggested that Pt-O v -Ti 3+ sites, in addition to Pt and TiO 2 sites, are active sites for a RWGS reaction [32]. The turnover frequency (TOF) of CO 2 conversion on reducible Pt/TiO 2 improves by approximately eight times compared with that on irreducible Pt/Al 2 O 3 ( Table 1) [32]. Kattel et al. [58] carried out DFT calculation and indicated that CO 2 adsorbs at the interface of Pt/TiO 2 (Fig. 2b); C adsorbs on Pt, and one O adsorbs on an oxygen vacancy (coordinatively unsaturated Ti cation), resulting in an enhanced CO 2 adsorption by 0.11 eV compared with that on Pt/SiO 2 . This process facilitates the cleavage of the C-O bond to form the adsorbed CO and fill the oxygen vacancy. Their experimental work has further confirmed that the stronger CO 2 adsorption on Pt/TiO 2 than that on Fig. 1 Sensitivity of CO and CH 4 production rate on the variation in a CO 2 binding energy [BE (CO 2 )] and b CO desorption energy on Pt nanoparticles [58] Pt/SiO 2 improves the TOF of CO 2 conversion in the RWGS reaction by 4-5 times (Table 1) [58]. Bobadilla et al. [62] reported that the activation energy of a RWGS reaction on Au/TiO 2 is 30 kJ/mol, which is much lower than that of 79 kJ/mol on Au/Al 2 O 3 ; as a result, the RWGS activity on Au/TiO 2 is higher than that on Au/Al 2 O 3 . As shown in Fig. 4, under identical reaction conditions (0.2 g of catalyst, H 2 /CO 2 = 4/1, gas hourly space velocity [GHSV] is 12,000 h −1 ), the conversion of CO 2 for Au/TiO 2 approaches the thermodynamic equilibrium of CO 2 conversion, while it is much lower on Au/Al 2 O 3 (Table 1) [62]. Bobadilla et al. [62] further performed DRIFTS and UV-vis spectroscopy studies and proposed that the presence of oxygen vacancies (or Ti 3+ cations) during the reaction facilitates the formation of a surface hydroxycarbonyl (OCOH) intermediate, which decomposes to form CO and H 2 O (Fig. 3b). CeO 2 is also a reducible support whose oxygen storage capacity (OSC) is higher than that of TiO 2 . CeO 2 -supported transition metal catalysts have been widely investigated for the mechanism of a RWGS reaction under various conditions [27,73,74]. An early work showed that CO 2 adsorption on Pt/CeO 2 at room temperature produces CO adsorbed on Pt and suggested that an oxygen vacancy in the vicinity of a Pt particle is responsible for CO 2 reduction to CO [75]. Performing DRIFT and mass spectroscopy (MS) during steady-state isotopic transient kinetic analysis, Goguet et al. [27] suggested that surface carbonate is an active intermediate, and formate is a spectator during the RWGS reaction on Pt/CeO 2 under their reaction condition. Wang et al. [74] conducted a quantitative transient analysis of products and showed that the surface oxygen vacancy in the vicinity of a Au particle in Au/CeO 2 can be replenished by an interaction with CO 2 and proposed that the RWGS reaction may follow a redox mechanism via the Au-assisted deposition and removal of active surface O. They also compared the OSC properties of Au/CeO 2 and Au/TiO 2 and suggested that oxygen vacancies away from an Au particle may be involved in the RWGS reaction on Au/CeO 2 , because of the higher oxygen mobility in CeO 2 than in TiO 2 . However, Chen et al. [38] combined Fourier transform infrared (FTIR) experiments and a temperature-programmed surface reaction and proposed that redox and associate mechanisms exist on a Pt/ CeO 2 catalyst via a related mechanism that involves a surface formate species as the major path under their reaction  [62] condition. These works have indicated that oxygen vacancies play an important role in the RWGS reaction, even though its reaction mechanism may vary depending on reaction conditions.
The RWGS activity of transition metals on CeO 2 is higher than that of metals supported on irreducible supports. Porosoff and Chen [76] compared the RWGS reaction on 1.7% Pt supported on reducible CeO 2 and irreducible Al 2 O 3 at 573 K (H 2 /CO 2 = 3/1, total pressure 30.0 Torr) and found that the forward reaction rate constant based on the mass of a catalyst and the amount of CO uptake (a measure of Pt metal dispersion) on Pt/CeO 2 are ~ 1.7 and ~ 5.7 times higher than those on Pt/Al 2 O 3 [76]. Yang et al. [63] prepared ~ 2 nm Cu supported on CeO 2 and mesoporous SiO 2 to exclude the effect of the particle size of Cu. They observed that Cu/CeO x is about four times more active than Cu/SiO 2 at 573 K (Table 1; 0.2 g of catalyst, H 2 /CO 2 = 3/1, 1 bar, weight hourly space velocity of 1.08 h −1 ) and assigned this improvement to the formation of oxygen vacancies on the Cu/CeO x catalyst.
In summary, the activity of metal catalysts on reducible supports (CeO 2 and TiO 2 ) in a RWGS reaction is higher than that on irreducible supports (SiO 2 and Al 2 O 3 ; Table 1), but they may form more CH 4 . No CH 4 forms on Au-and Cu-based catalysts, whereas CH 4 may form on Pt and Pd catalysts [62,63]. For Pt catalysts, CH 4 selectivity is higher on reducible supports (TiO 2 and CeO 2 ) than on irreducible supports (SiO 2 and Al 2 O 3 ) possibly because of the removal of both O atoms from CO 2 by oxygen vacancies and the hydrogenation of the formed C species to CH 4 [76]. The deposited carbon species on a support may lead to catalyst deactivation. Indeed, Goguet et al. [73] reported the deactivation of a 2% Pt/CeO 2 catalyst during a RWGS reaction. They subjected the used catalysts to temperatureprogrammed oxidation and found that deactivation is closely related to the amount of deposited carbon. Thus, carbon species deposited on partially reduced CeO 2 around a Pt particle is the major reason for this deactivation.
Although metal/reducible support catalysts show higher activity in RWGS, the work of direct comparison between different reducible supports and their consequences on the varied activity and selectivity for RWGS reaction is rare. Sakurai et al. [64] studied CO 2 hydrogenation on Au/CeO 2 , Au/ZrO 2 , and Au/TiO 2 catalysts. Under identical reaction conditions [250 °C, H 2 /CO 2 = 3, 8 atom pressure, 3000 mL/ (g cat h)], these catalysts show similar CO selectivity (> 90%), whereas conversion is quite different: 3.4%, 9.2%, and 16.1% for Au/CeO 2 , Au/ZrO 2 , and Au/TiO 2 , respectively ( Table 1). This high activity on Au/TiO 2 is attributed to the higher acidity of TiO 2 than those of ZrO 2 and CeO 2 . Kattel et al. [65] compared CO 2 hydrogenation on bimetallic PtCo supported on CeO 2 , ZrO 2 , and TiO 2 . Similar CO 2 conversion (7.8%-9.1%) is observed on these catalysts under identical condition (0.1 g of catalyst, 300 °C, H 2 /CO 2 = 2/1, feed flow rate of 60 mL/min), whereas the highest TOF and the highest CO selectivity are found on a PtCo/TiO 2 catalyst (Table 1). These ascribed this change to the strength of C, O-bound, and O-bound species are enhanced on PtCo/ CeO 2 and PtCo/ZrO 2 interfaces with respect to that of PtCo/ TiO 2 . As a result, different reaction intermediates form, and CH 4 selectivity increases. We note that the surface area of these reducible supports are different, perhaps resulting in a comparison with different bases. Therefore, how the redox and acid-base properties of different supports influence the activity and selectivity of RWGS reaction should be further studied.

Tuning Metal-Support Interactions
A RWGS reaction occurs at the interface of a metal and a support, so finely tuning metal-support interactions and maximizing metal-support interfacial sites can greatly improve activity and enhance selectivity toward CO. Maximizing metal-support interfacial sites can be achieved by reducing the particle size of metals, adjusting pretreatment procedures, and tuning the property of supports. Tuning metal-support interactions is particularly important for Ni-, Rh-, and Ru-based catalysts, which are more selective to the production of CH 4 rather than CO.
Reducing the particle size of transition metals not only provides more surface sites for a reaction but also increases the number of metal-support interfacial sites and has been investigated by several groups. Kwak et al. [66] tested CO 2 hydrogenation on 0.1%-5% Ru/Al 2 O 3 catalysts. CH 4 is the Fig. 4 The catalytic performance of Au/TiO 2 and Au/Al 2 O 3 catalysts in the RWGS reaction. Reaction conditions: 0.2 g catalyst, H 2 / CO 2 = 4/1, feed flow rate 50 mL/min, and gas hourly space velocity (GHSV) from 12,000 to 40,000 h −1 [62] major product on a 5% Ru sample (mainly Ru particles), while CO is the major product on a 0.1% Ru sample (mainly atomically dispersed Ru; Table 1). On the 0.1% Ru/Al 2 O 3 catalyst at 350 °C (Fig. 5), CO is the only product at the beginning of a reaction, whereas CH 4 becomes the major product with time on stream. This is because the sintering of atomically dispersed Ru on particles occurs with time on stream possibly due to a weak Ru-Al 2 O 3 interaction. Matsubu et al. [56] prepared 2% Rh/TiO 2 catalyst with both Rh particles and isolated Rh sites. They removed the Rh particles without removing isolated Rh species that strongly interacts with TiO 2 by leaching with HCl/H 2 O 2 (Fig. 6a). In comparison with a fresh Rh/TiO 2 catalyst with Rh particles exhibiting the major product of CH 4 , the leached Rh/ TiO 2 catalyst with only the isolated Rh sites produces CO at 200 °C and CO 2 /H 2 of 3 (15-30 mg of catalyst, 0.1 MPa, CO 2 /H 2 /N 2 = 3/1/96, feed flow rate of 100 mL/min; Fig. 6). Therefore, they concluded that an isolated Rh site is the active site for a RWGS reaction, while Rh particle is the active site for CH 4 formation. Combining DFT and experiments, Chen et al. [77] also demonstrated that a single Ir atom interacting with TiO 2 is selective for CO formation, whereas Ir particle is active for CH 4 formation. These works have mainly focused on the particle size of metals but have rarely explored the role of metal-support interactions.
However, one may expect that metal-support interactions can be enhanced by decreasing the particle size of metals because of an increased metal-support interface. Recently, Li et al. [67] prepared 5% Ir/CeO 2 catalyst with an Ir particle size of approximately 1 nm and 0.7% Ir/CeO 2 with atomically dispersed Ir. Although an Ir/CeO 2 catalyst with large Ir particles (> 2.5 nm) shows the dominant product of CH 4 , the 5% and 0.7% Ir/CeO 2 catalysts have the dominant product of CO. The activity based on the mole of Ir improves when the size of an Ir particle is smaller than 1 nm (Table 1; 300 °C, 1 MPa, H 2 /CO 2 /Ar = 76/19/5). Li et al. [67] also conducted X-ray adsorption spectroscopy (XAS) and found that 1 nm Ir particle and atomically dispersed Ir become partially oxidized because of the strong interaction with CeO 2 , whereas a large Ir particle (> 2.5 nm) is mainly reduced. Thus, they argued that the main active site for a RWGS reaction is the partially oxidized Ir that strongly interacts with CeO 2 support irrespective of Ir particles or atomically dispersed Ir species.
Varying treatment conditions can tune the metal-support interactions of metal/reducible support catalysts. Aitbekova et al. [55] prepared ~ 2.6 nm Ru uniformly dispersed on Al 2 O 3 , TiO 2 , and CeO 2 supports and tested in a CO 2 reduction reaction. They found that TiO 2 -and CeO 2 -supported Ru are more active than Al 2 O 3 -supported Ru catalyst and that CH 4 is the major product in all catalysts. However, the mild oxidation of catalysts at 230 °C and followed by lowtemperature (230 °C) reduction results in the redispersion of Ru particles to form a single RuO x site on CeO 2 , which causes an almost complete shift of the product selectivity from CH 4 to CO (Fig. 7). This enhanced selectivity is attributed to the weakened adsorption of CO on the single RuO x site. However, this redispersion partially occurs in Al 2 O 3and TiO 2 -supported Ru possibly because of the weaker interaction of Al 2 O 3 and TiO 2 than that of CeO 2 with RuO x to disperse the single RuO x site. The mild oxidation at 230 °C and followed by the high-temperature (500 °C) reduction of Ru/TiO 2 favor the formation of a strong metal-support interaction (SMSI). That is, the partially reduced TiO x covers Ru particles, resulting in a decreased CO 2 conversion and an enhanced CO selectivity due to the loss of a surface Ru site by covering. Interestingly, the redispersed Ru/CeO 2 catalyst via a mild oxidizing treatment is stable in a RWGS reaction for 12-14 h, and CO selectivity increases further from 90% to 99%. Similarly, the high-temperature (500 °C) reduction of a Ni/TiO 2 catalyst results in a SMSI effect with a decreased activity, but this reaction shifts the selectivity from CH 4 to CO (Table 1) because Ni is covered, and more Ni-TiO 2 interface sites are produced [57]. However, for the supported Ni catalysts prepared from NH 3 or H 2 -pretreated TiO 2 support, reduction at 500 °C only induces a SMSI effect to a limited extent; thus, CH 4 is the dominant product on these catalysts (Table 1).
In addition to changing metal particle sizes or pretreatment conditions, varying the property of a support can also be used to tune metal-support interactions. Kim et al. [78] prepared a series of Pt/TiO 2 catalysts with a varying crystallite size of TiO 2 . They found that decreasing the crystallite size of TiO 2 significantly improves the reducibility of Pt/ TiO 2 , produces more Pt-Ov-Ti 3+ interfacial sites, and eventually improves the activity in a RWGS reaction (Fig. 8).

Addition of Reducible Transition Metal Oxide Promoters
The addition of a promoter can also change the adsorption and activation of CO 2 and may further change the activity and selectivity for a RWGS reaction. One type of a promoter is reducible transition metal oxide, which can also form oxygen vacancies upon reduction. Similar to oxygen vacancies on reducible TiO 2 and CeO 2 , oxygen vacancies at the interface of a metal-reducible metal oxide promoter may play an important role in CO 2 activation. Adjusting the interaction   [78] between a metal and a promoter may significantly improve the activity and tune selectivity in a RWGS reaction.
Kwak et al. [61] compared the CO 2 reduction on 1% Pd/ MWCNT and 1% Pd-2.3% La 2 O 3 /MWCNT catalysts (0.05 g of catalyst, H 2 /CO 2 /He = 3/1/16, feed flow rate of 60 mL/ min). In contrast to the inactivity of 1% Pd/MWCNT for CO 2 reduction, addition of La 2 O 3 to the catalyst improves the activity, and CO is observed as the only product (< 500 °C; Table 1). This observation highlights the role of La 2 O 3 in CO 2 activation. Ro et al. [79] prepared a MoO x -modified Pt/SiO 2 catalyst via a controlled surface reaction. MoO x is selectively deposited on surface Pt at a Mo/Pt ratio of < 0.3, thereby creating Pt-MoO x interfacial sites. The intrinsic reaction rate and TOF of the optimized PtMo/SiO 2 (Mo/ Pt = 0.3) for the RWGS reaction at 473 K improve by ~ 11 and ~ 39 times compared with those of Pt/SiO 2 (Fig. 9), respectively (H 2 /CO 2 = 2, 7.1 bar). The activation energy decreases from 67.9 kJ/mol for Pt/SiO 2 to 60.0 kJ/mol for PtMo/SiO 2 . The reaction orders are altered upon MoO x loading. Interestingly, the activity of the RWGS reaction can be further improved when Pt-MoO x is exposed to light irradiation. Yan et al. [68] investigated CO 2 reduction on NiFe/ ZrO 2 catalysts with varying amounts of Fe loadings. As shown in Fig. 10, at low Fe loadings (Ni/Fe > 1), the major active site is the Ni-ZrO 2 interface, which selectively produces CH 4 (selectivity > 88%) through methanation (673 K, H 2 /CO 2 /Ar = 2/1/5). At high Fe loadings (Ni/Fe = 1/3), the major active site is a Ni-FeO x interfacial site, producing CO with selectivity higher than 95% (Table 1). DFT calculations have indicated that the CO adsorption energies at Ni-ZrO 2 and Ni-FeO x interfaces are −2.92 and −1.93 eV, respectively. Therefore, CO adsorbed at the Ni-ZrO 2 interface is too strong, so it favors CH 4 formation. Conversely, CO adsorbed at the Ni-FeO x interface is weaker, so it promotes CO desorption.

Bimetallic Alloy Formation
Another type of promoter can be completely reduced to form a bimetallic alloy with a major metal [80][81][82]. Depending on the property of promoters, bimetallic alloy properties, such as H 2 dissociation, CO 2 adsorption, and C-O cleavage, may be changed. For example, Yuan et al. [81] performed a DFT study on CO 2 hydrogenation on Ni (111) and Re-doped Ni (111) surfaces. The adsorption energies of CO 2 on Ni (111) and Re@Ni (111) surfaces are 0.20 and −0.28 eV, respectively. This result indicates the stronger adsorption of CO 2 on the latter surface. Furthermore, the activation energy of CO 2 dissociation to CO* and O* decreases from 0.41 eV on Ni (111) to 0.18 eV on Re@Ni (111). These results suggest that the presence of oxophilic Re on Ni (111) favors one O atom of CO 2 adsorption on Re and promotes the C-O bond dissociation. Thus, a bimetallic alloy may also change the activity and selectivity of a RWGS reaction.
Alayoglu et al. [80] showed that Pt is segregated on the surface of a PtCo alloy, and the selectivity of CO increases from ~ 82% on Co/MCF-17 to ~ 100% on PtCo/MCF-17 at a CO 2 conversion of ~ 5% (reaction conditions: 0.05 g of   [68] catalyst, 300 °C, 5.5 bar, H 2 /CO 2 /He = 33.3/11.1/5.6, feed flow rate of 50 mL/min). Porosoff and Chen [76] compared CO 2 reduction on bimetallic PtCo, PtNi, and PdNi supported on Al 2 O 3 or CeO 2 catalysts. PdNi/CeO 2 is the most active catalyst, but it forms the highest amount of CH 4 . By comparison, PtCo/Al 2 O 3 shows the highest selectivity toward CO. They suggested that selectivity can be correlated with the electronic properties of catalysts by using values of the d-band center. As shown in Fig. 11, the Pt-Co bimetallic catalyst with a d-band center value away from the Fermi level and toward the more negative direction yields the highest CO:CH 4 ratio. The advantage of establishing this trend is that it can provide a direction for the future selection of single-metal or bimetallic catalysts that efficiently activate CO 2 . Further studies have indicated that PtCo/TiO 2 is more selective than PtCo/CeO 2 and PtCo/ZrO 2 catalysts for CO because of the weak bonding strength of CO at the Pt-Co/ TiO 2 catalyst interface site [65]. Ye et al. [59] compared CO 2 reduction on Pd/SiO 2 and PdIn/SiO 2 catalysts. A Pd/ SiO 2 catalyst is much more active than other catalysts, but it produces CH 4 as a minor product. By comparison, PdIn/ SiO 2 is much less active than other catalysts, but it only produces CO (Table 1). Characterizations have shown the formation of a PdIn bimetallic alloy (Fig. 12). A DFT study has revealed that a PdIn alloy is less active than Pd for H 2 dissociation, CO adsorption on the PdIn alloy is weaker than that on Pd, and PdIn is energetically less favorable to the hydrogenation of CO to CH 4 than Pd. Specifically, the H 2 dissociation energy on a PdIn alloy increases by 1.56 eV, whereas the linear CO adsorption energy on a PdIn alloy decreases by 0.77 eV. These changes shift the formation of CH 4 on Pd/SiO 2 to the formation of CO on bimetallic PdIn/ SiO 2 .

Addition of Alkali Metals
The addition of alkali metals to Pt-and Au-based catalysts can greatly improve the activity of a water gas shift (WGS) reaction [83][84][85]. The addition of an appropriate amount of Na to Pt/TiO 2 can improve reducibility and surface basicity and form strong Pt-O-Na interactions by donating electrons from Pt to NaO x , which creates an active site at the interface of Pt-NaO x for the WGS reaction [83]. The increased basicity may improve the adsorption of acidic CO 2 . Furthermore, electronically modified Pt may alter interactions with CO, enhance activity, and tune the selectivity for the RWGS reaction.
Chen et al. [69] showed that the addition of a small amount of K 2 O to a Cu/SiO 2 catalyst improves its activity for a RWGS reaction by more than 2.4 times (Table 1). Specifically, the CO 2 conversion is 12.8% on a Cu/K 2 O/ SiO 2 catalyst, while it is only 5.3% on Cu/SiO 2 catalyst at 600 °C (0.02 g of catalyst, 0.1 MPa, H 2 /CO 2 = 1/1, feed flow rate of 100 mL/min). K likely enhances CO 2 adsorption and facilitates the decomposition of the formate species to CO. Liang et al. [70] studied the effect of the addition of 2% K to 2% Pt/mullite on the RWGS reaction. The TOF of RWGS at 340 °C improves by seven times with the addition of K (Table 1; H 2 /CO 2 /N 2 = 4.5/4.5/1, feed flow rate of 50 mL/ min) with a reduced selectivity to CH 4 (Fig. 13). The activation energy of CO 2 conversion decreases from 62 ± 2.3 kJ/ mol to 34 ± 1.3 kJ/mol upon K addition. This result suggests that the interface between KO x and Pt serves as an active site for the decomposition of formate to CO, and the presence of KO x weakens CO adsorption on Pt and thus hinders CH 4 formation. Yang et al. [86] investigated the role of K addition in the activity and selectivity of Pt on zeolite L (Pt/L) for a RGWS reaction. K addition (K/Pt = 80) improves the reaction rate from 0.021 mol/(g cat h) to 0.080 mol/(g cat h) at 500 °C compared with that of Pt/L (0.1 g of catalyst, H 2 / CO 2 /N 2 = 4.5/4.5/1, feed flow rate of 50 mL/min), while CH 4 formation decreases. On the basis of XAS and X-ray photoelectron spectra (XPS) results, Yang et al. [86] proposed the formation of a Pt-O(OH)-Pt interface, which serves as an active site to adsorb CO 2 and form an active intermediate of bridge-bonded formate. Similarly, Santos et al. [84] reported that the addition of Na to Pt/C can improve the RWGS reaction with a reduced activation energy. They suggested that the addition of Na modifies the electronic structure of Pt and favors CO 2 dissociation.
In addition to improving the activity of the RWGS reaction, changing product selectivity can be achieved by adding alkali metals. Bando et al. [71] investigated the effect of Li modification on Rh-exchanged Y zeolite for CO 2 hydrogenation (1 g of catalyst, 250 °C, 3 MPa, H 2 /CO 2 = 3/1, feed flow rate of 100 mL/min). They found that CH 4 is the major product of Rh-Y, whereas adding Li to Rh-Y mainly Fig. 11 Effect of the d-band center on the ratio of CO to CH 4 production at 10% conversion of the RWGS reaction. Open and solid symbols represent catalysts with and without Ni, respectively. Reaction conditions: 573 K, 0.004 MPa, H 2 /CO 2 = 3/1 [76] produces CO (Table 1). This change is attributed to Li that provides new active sites for CO 2 adsorption and stabilizes the adsorbed CO.

Enveloping Metal Particles
These above strategies are based on tuning metal-support (promoter) interactions. The local environment of metal particles is also important for various reactions and may be finely tuned by enveloping with nanoporous materials. Wang et al. [72] reported a new strategy to tune the local environment of metals by enveloping a metal particle with different zeolites for CO 2 reduction reaction. The Rh particles of 4.2-5.0 nm are either supported on zeolite (Rh/ HZSM-5, Rh/S-1) or enveloped in HZSM-5, KZSM-5, and pure silica zeolite of S-1 (Rh@HZSM-5, Rh@KZSM-5, and Rh@S-1). CO 2 reduction reaction tests (0.5 g of catalyst, 1 MPa, H 2 /CO 2 /Ar = 3/1/1, feed flow rate of 30 mL/min, 250-500 °C) showed that Rh/HZSM-5, Rh/S-1, and Rh@ HZSM-5 favor CH 4 formation, whereas Rh@S-1 produces CO as a major product (Table 1) and CH 4 as a minor product (Fig. 14). The selectivity behavior of Rh particles enveloped in S-1 is similar to that of the single Rh site on TiO 2 [56]. FTIR and H 2 spillover experiments showed that proton form HZSM-5 favors stronger CO adsorption on Ru and facilitates hydrogen spillover, while pure silica zeolite favors weaker CO adsorption and limited hydrogen spillover. The weak CO adsorption favors desorption, and the limited hydrogen spillover decreases the Rh hydrogenation ability toward the CH 4 formation. A stability test at 400 °C has demonstrated that Rh/S-1 is stable for 150 h with a CO selectivity of 96%.

Conclusions and Outlooks
The catalytic conversion of CO 2 to CO via the RWGS reaction followed by well-established synthesis gas conversion technologies may provide a potential approach to convert CO 2 to valuable chemicals and fuels. The improvement in the low-temperature activity and selectivity of the RWGS reaction is a key challenge for this reaction. We reviewed recent advances in the strategies of catalyst designs that can  [59] enhance the activity and selectivity of the RWGS reaction. These strategies include varying support, tuning metal-support interactions, adding reducible transition metal oxide promoters, forming bimetallic alloys, adding alkali metals, and enveloping metal particles. The analysis of these advances suggests that enhancing CO 2 adsorption and facilitating CO desorption appear to be the key factors for enhancing CO 2 conversion and CO selectivity.
Although significant efforts have been made, studies have shown that the RWGS reaction is accompanied with a methanation side reaction on various catalysts. As a result, a mixture of CO and CH 4 products forms, thereby burdening subsequent separation processes. Thus, RWGS catalysts with 100% CO selectivity should be designed. In the future, combining different metals with supports/promoters and tailoring the fine structure of supported metal catalysts should be considered the important directions for designing highly selective catalysts of the RWGS reaction.
Future work may focus on the reaction mechanism of catalysts in a RWGS reaction. The reaction mechanism and even active sites have been strongly debated, and these aspects are essential for the rational design of highly efficient and selective RWGS catalysts. With the development of in situ and operando spectroscopy technologies, such as in situ XPS and in situ XRD, as well as the capability of DFT calculations, the reaction mechanism on various catalysts operating under various reaction conditions may be clarified in the future. With better understanding on the reaction mechanism and structure-activity/selectivity relationship, more efficient catalysts may be designed in the future.
Future work may also focus on the stability of catalysts in a RWGS reaction. The stability of catalysts in this reaction has been rarely explored. Several stability tests have been performed under conditions near the thermodynamic equilibrium of CO 2 conversion, which may mask the unstable nature of catalysts. The stability of catalysts is an important issue for a commercial process, so this parameter should be extensively investigated. Understanding the deactivation mechanism may also provide useful information for the rational design of stable catalysts in the RWGS reaction. Fig. 13 Comparison of the turnover frequency (TOF) and the catalytic activity of RWGS reaction on Pt/mullite and Pt-K/mullite catalysts. Reaction conditions: 0.1 g of catalyst, 340 °C, 0.1 MPa, H 2 /CO 2 /N 2 = 4.5/4.5/1, feed flow rate of 50 mL/min [70] Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creat iveco mmons .org/licen ses/by/4.0/.